K-most-interesting rule discovery allows a user to discover interesting rules from data, where the user is empowered to define the criteria by which a rule is judged interesting. This project will investigate new criteria of interestingness and develop techniques to support rule discovery in the context of those criteria.
Degree Suitability: Computer Science, Digital Systems, or Software Engineering
References:
G. Webb (2000) Efficient search for association rules. Proceedings of KDD-2000: the SIGKDD Conference of Knowledge Discovery and Datamining, Boston, pp.99-107.
G. Webb (1995) OPUS: An efficient admissible algorithm for unordered search. Journal of Artificial Intelligence Research., 3: 431-465.
There are relatively few data mining techniques for analysis of quantitative (numeric) data. However, much data is quantitative, and there is a tremendous unmet demand for techniques that support sophisticated quantitative data analysis. This project will investigate extensions to the impact rules technique for learning rules that identify interesting distributions on a numeric variable.
Degree Suitability: Computer Science, Digital Systems, or Software Engineering
References:
G. Webb (2001) Discovering associations with numeric variables. Proceedings of
KDD-2001: the SIGKDD Conference of Knowledge Discovery and Datamining, San Francisco,
pp.383-388.
G. Webb (1995) OPUS: An efficient admissible algorithm for unordered search. Journal of
Artificial Intelligence Research., 3: 431-465.
AODE is an efficient new technique for estimating conditional probabilities from data that has demonstrated high levels of classification accuracy. This project will investigate extensions to AODE that seek to further improve AODE's efficiency while retaining its prediction accuracy.
Degree Suitability: Computer Science, Digital Systems, or Software Engineering
References:
G. Webb, J. Boughton & Z. Wang (2002) Averaged one-dependence estimators:
Preliminary results. Proceedings of the Australasian Data Mining Workshop, Canberra.
Z. Zheng & G. Webb (2000) Lazy learning of Bayesian rules. Machine Learning, 41(1):
53-84.
It is relatively easy for a machine learning system to learn accurate classifiers for objects that are highly similar to those in the training data. However, the prediction accuracy of many systems falls as the similarity of test objects to training objects decreases. This project will investigate this effect and develop new learning techniques that seek to minimise it.
Degree Suitability: Computer Science, Digital Systems, or Software Engineering
Many computer systems can benefit from anticipating a user's likely intentions. This project will extend the Feature-Based Modelling approach to developing models of a user's interactions with a computer system.
Degree Suitability: Computer Science, Digital Systems, or Software Engineering
References:
G. Webb, M. Pazzani, D. Billsus (2001) Machine learning for user modeling. User
Modeling and User-Adapted Interaction, 11(1-2): 19-29. Tenth anniversary issue.
B. C. Chiu & G. Webb (1998) Using decision trees for agent modelling: Improving
prediction performance. User Modeling and User-Adapted Interaction, 8(1-2): 131-152.
G. Webb, B. C. Chiu & M. Kuzmycz (1997) Comparative evaluation of alternative
induction engines for Feature Based Modelling. International Journal of Artificial
Intelligence in Education, 8: 97-115.
G. Webb & M Kuzmycz (1996) Feature Based Modelling: A methodology for producing
coherent, consistent, dynamically changing models of agents' competencies. User Modeling
and User-Adapted Interaction. 5 (2): 117-150.